What the Job Change Tool Actually Fixes on a 2-Furnace Plant

Thirty to sixty percent. That is the cross-shift variance you will find on identical SKUs at plants that have not systemised their job change process. Not plant-to-plant variance. Section-to-section, shift-to-shift, on the same forming line, running the same bottle. Most plant managers I speak with have never measured this number. They know their OEE. They know their cullet percentage. The job change variance is in someone's notebook, if it's anywhere at all.

That is the problem the systemised Job Change Tool was built to address. Not to make changeovers faster for the sake of a metric on a whiteboard. To stop the same SKU performing differently depending on who happened to be on shift.

The gap between your best and worst shift is not a people problem

On a 2-furnace plant running five IS lines, you will average two or three job changes per week across the plant. Each one resets your forming conditions from scratch. Moulds out, moulds in, new recipe loaded or recalled from memory, and then somewhere between mould change completion and first-ware inspection, every operator on every section is making judgement calls. That is where the variance lives.

In 2016, running a plant in regional Australia, we pulled the job change records for one of our 375ml amber wine bottles. Same moulds, same glass, same IS machine. A Lynch 10-section running late-1990s controls. The first-ware reject rate swung from 8% on the best shift to 31% on the worst, across the same SKU. Twenty-three minutes. That was the average difference in changeover time between those two shifts.

Nobody was doing anything technically wrong. They just weren't doing the same things in the same sequence with the same targets. Not a performance management issue. A system absence.

The gap between your best shift and your worst on the same SKU is not a people problem. It is a system problem. And system problems need system fixes, not more conversations at shift handover.

What the SKU Library actually does on a forming line

The SKU Library is the first component of the Job Change Tool, and it's the one that sounds least interesting right up until you've watched a night-shift operator reconstruct a recipe from a WhatsApp message because the previous version lived on a supervisor's laptop that was replaced eight months ago.

Every SKU in the library carries the full forming spec: gob weight target and acceptable CV (the working standard for lightweight containers is ≤0.4%), section timing to within 10ms, mould preheat curve target (480°C ±10°C as a baseline, adjusted for glass composition and container geometry), plunger stroke, and cooling profiles. Version-locked. The hot-end superintendent owns recipe lock. The operator does not change set points without sign-off. That boundary matters more than most plants enforce it.

And this matters most at the 0600 handover, which is the most information-leaky moment in the operating cycle. Night shift knows what they changed, why they changed it, and what the ware looked like at T-30 min before handover. Day shift inherits the outcome with none of that context. A locked, versioned SKU record doesn't replace the conversation. It gives the conversation something reliable to stand on.

Live Execution: a checklist that knows where you are in the job

The second component maps to the 9-stage Job Change Lifecycle: plan, prep, line-down, mould change, recipe load, ignition, first ware, stabilise, post-mortem. Each stage has named owners and a completion gate before the next opens. The hot-end superintendent owns plan approval and recipe load sign-off. QA owns first-ware acceptance. The mould shop owns mould change readiness. Not shared ownership. Named ownership.

The operator-facing view is a section-by-section checklist. When section 4 completes recipe load, it shows complete. Section 7 still mid-mould-change? It stays incomplete. The superintendent can see that split in real time, during the job, not at end-of-shift when the paperwork arrives. That visibility moves the decision point upstream, which is the only place where decisions are still cheap.

Baffle marks at first ware usually point at baffle alignment drift carried over from the previous run. Checking the centring tool reading first (and yes, I know your fitter says it's fine, check it anyway) is a five-minute task at mould change. Not catching it there means catching it at first-ware inspection, which is a different conversation and a longer one. Live Execution pushes that check to where it belongs in the sequence.

KPI Tracking: the data that makes the post-mortem worth running

Most plants I've audited track OEE, tonnes packed, and cullet percentage. Some track changeover time as a single plant-wide number. Almost none track job change performance at section level, across shifts, trended over time per SKU. Without that data, the post-mortem stage of every job change is four minutes of conversation and no evidence.

The Job Change Tool captures four things at each changeover:

• Changeover time, broken out by section and in total
• First-ware quality result against the SKU acceptance spec
• Time-to-stable-pack from the point of line-up
• Section-level variance against the recipe targets

What that gives a plant manager is a diagnostic layer, not just a record. If section 6 consistently adds 18 minutes to every job change across three different operators, that is a section-side mechanical issue, not an operator issue. If one shift reaches stable-pack 40 minutes later than the other two shifts on the same SKU, that is either a recipe execution gap or a handover gap. The data tells you which.

That feedback loop is where the long-run OEE gain sits. Most container glass plants leave 4-8 OEE points on the floor through uncontrolled changeover variance, and it is the most controllable category of loss on the hot end. In European plants operating under EU ETS Phase IV, where every rejected tonne carries the full melt energy cost, that inefficiency compounds at the margin level much faster than it does in plants with looser carbon exposure.

The three components work as a system. The SKU Library sets the standard. Live Execution makes the standard repeatable. KPI Tracking tells you whether it's holding and where it's drifting. A vendor-neutral review of your current job change performance is often the fastest way to identify which component is weakest in your operation. If you want to understand how the SMED framework applies to IS forming in practice, the entry on SMED for container glass is worth reading first. An independent container glass consultant will tell you the same thing the data eventually does: generic SMED doesn't account for section-level variance, and that's where most of the controllable loss is hiding.